How to diagnose cause of singular matrices?

[JJ Hubbard]

I am fitting an AR model like so:

m1 <- inla(Y ~ 1 +

    # Autoregressive

    + f(time_step, model="ar",

        order=1,

        replicate=participant_f,

        values=1:max(df$time_step),

        hyper = hyper_participant_f,

        constr=FALSE

    )

           

    , data=df, family="gaussian"

    , control.inla=list(strategy="adaptive",  int.strategy="auto")

    , control.compute=list(return.marginals.predictor=TRUE, cpo=TRUE, waic=TRUE, dic=TRUE, config=TRUE)

    , verbose=TRUE

    )

But INLA is frequently crashing due to the Hessian becoming singular and all NaN. What is the procedure for investigating what makes this matrix so easy to become singular?

Thank you very much.

[INLA help]

Can you retry with adding

inla(…., inla.mode=‘experimental’)

Get Outlook for iOS

---

From: r-inla-disc...@googlegroups.com <r-inla-disc...@googlegroups.com> on behalf of JJ Hubbard <jordanj...@gmail.com>
Sent: Tuesday, August 16, 2022 7:15:20 PM
To: R-inla discussion group <r-inla-disc...@googlegroups.com>
Subject: [r-inla] How to diagnose cause of singular matrices?

 

--
You received this message because you are subscribed to the Google Groups "R-inla discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to r-inla-discussion...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/r-inla-discussion-group/5e6486fb-0172-44f7-995a-b5f493520a9dn%40googlegroups.com.

[JJ Hubbard]

I have been running in experimental mode. Switching from stable to testing version seems to have made a massive improvement.

[INLA help]

Add verbose=T and look at the parameter in the optimization.  I guess one of them go high.  The name of the parameters are given when Optimization starts 

Get Outlook for iOS

---

From: r-inla-disc...@googlegroups.com <r-inla-disc...@googlegroups.com> on behalf of JJ Hubbard <jordanj...@gmail.com>

Sent: Tuesday, August 16, 2022 7:54:58 PM

To: R-inla discussion group <r-inla-disc...@googlegroups.com>

Subject: Re: [r-inla] How to diagnose cause of singular matrices?

 

To view this discussion on the web, visit https://groups.google.com/d/msgid/r-inla-discussion-group/a4abb83b-9f44-4183-b18d-2b99a29a4d2bn%40googlegroups.com.